Antibiotic therapy of experimental Pseudomonas keratitis was evaluated quantitatively by determining numbers of viable bacteria in the cornea of guinea pigs. Topically applied carbenicillin disodium, gentamicin sulfate, and tobramycin sulfate were often significantly more effective than topically applied polymyxin B sulfate. Intramuscular therapy with tobramycin was as effective as topical therapy, and the results exhibited less variability. Topical tobramycin every 30 minutes was significantly more effective than topical therapy every 60 minutes. No combination of antibiotics was significantly better than a single effective drug. The concentration of tobramycin in the aqueous correlated more closely to therapeutic efficacy than did the concentration in the cornea. Although all antibiotics reduced numbers of bacteria in the cornea by more than 99% in the first 24 hours of therapy, none was able to sterilize the cornea in four additional days of continuous therapy. Persistence of organisms despite apparently adequate topical therapy may explain some reported cases of relapse in humans.